Skip to main content

​
Overview

Retrieve element information in two modes:
  • Semantic mode (default): 70-99% token reduction, returns [Role] "Name" (properties) format ideal for AI agents
  • Raw mode (--raw): Full HTML with attributes for debugging and inspection

​
Syntax

bdg dom get <selector> [options]

​
Arguments

​
selector
string
required
CSS selector or numeric index from cached query results (0-based)

​
Options

​
--raw
flag
Output raw HTML instead of semantic representation
​
--all
flag
Get all matching elements (only with --raw)
​
--nth
number
Get nth matching element, 0-based (only with --raw)
​
--node-id
number
Use CDP nodeId directly (only with --raw)
​
--json
flag
Output as JSON

​
Output modes

​
Semantic mode (default)

Returns accessibility representation with role, name, and key properties: 
$ bdg dom get "h1"
[Heading L1] "Welcome to the Dashboard"

$ bdg dom get "button"
[Button] "Submit Form" (focusable)

$ bdg dom get "#searchInput"
[Searchbox] "Search" (focusable, required)
Token efficiency:
  • Simple elements: 45-75% reduction
  • Complex elements: 82-99% reduction
  • No CSS classes or non-semantic attributes

​
Raw HTML mode

Returns complete HTML with all attributes: 
$ bdg dom get "button" --raw
<button type="submit" class="btn btn-primary" id="submit-btn" aria-label="Submit Form">
  Submit Form
</button>

​
Examples

​
Get semantic representation

# By selector
bdg dom get "button.primary"
# [Button] "Submit" (focusable)

# By cached index
bdg dom query "button"
bdg dom get 0
# [Button] "Submit" (focusable)

​
Get raw HTML

# First match
bdg dom get "button" --raw

# All matches
bdg dom get "button" --raw --all

# Specific match by index
bdg dom get "button" --raw --nth 2

# By CDP nodeId
bdg dom get --raw --node-id 123

​
Use with cached query results

# Step 1: Query and cache
bdg dom query ".nav-link"
# Found 5 elements

# Step 2: Get details by index
bdg dom get 0  # First nav link
bdg dom get 4  # Last nav link

​
Semantic output format

​
Role text

  • Heading levels: [Heading L1], [Heading L2], etc.
  • Standard roles: [Button], [Link], [Textbox], etc.
  • First letter capitalized

​
Context

  • If accessible name exists: "Name"
  • Fallback to DOM context: <tag.class> "text preview"

​
Properties

  • focusable - Element can receive keyboard focus
  • focused - Element currently has focus
  • disabled - Element is disabled
  • required - Field is required
  • (inferred from DOM) - Role synthesized from HTML

​
Examples

[Heading L1] "Welcome"                          # Heading with name
[Button] "Submit" (focusable)                   # Focusable button
[Link] "Learn more" (focusable)                 # Focusable link
[Searchbox] "Search" (focusable, required)      # Required searchbox
[Textbox] <input.email> "user@" (focusable)     # Input with partial value
[Navigation] "Main menu"                        # Navigation landmark
[Paragraph]                                      # Paragraph without text

​
When to use --raw

Use raw HTML mode when you need:
  • Exact HTML structure with classes and attributes
  • Multiple elements (--all)
  • Specific element selection (--nth, --node-id)
  • CSS/HTML debugging
  • Data attributes or custom properties
Use semantic mode (default) for:
  • AI agent automation
  • Token-efficient context
  • Accessibility-first element identification
  • Form field discovery

​
Exit codes

​
0
number
Success - element retrieved
​
83
number
RESOURCE_NOT_FOUND - element not found
​
87
number
STALE_CACHE - cached index invalid (page navigation)
​
102
number
CDP_TIMEOUT - CDP operation timed out

Build docs developers (and LLMs) love

Get started for freeTalk to us